MAE Testing the Treequest Search

After @will ran the pipeline, we are left with a handful of our best candidates to continue validating. The next filter they need to pass is a decent magnetocrystalline anisotropy energy.

Check out Will's writeup on the methodology:

For reference, the MAE model predicts FePt at 3.07 meV. This is really good! I haven't directly tested Nd₂Fe₁₄B but literature reports values around 2.9 meV per unit cell. Anything above ~2.5 is likely to be a good candidate! Most materials are less than 0.1 meV so when you have values of multiple meV that's a solid material.

Now let's take a look at some of the top candidates:

• Number of atoms in the cell: 4

• Magnetization density: 0.118825

• e_above_hull: 0.066408 eV/atom

• Curie temperature: 672.8 K

• No imaginary modes detected

• Easy axis z: -9789.619323 eV (θ=0.0°, φ=0.0°)

• Hard axis x: -9789.619306 eV (θ=90.0°, φ=0.0°)

• ΔMAE (hard-easy): 0.016494 meV

• Number of atoms in the cell: 4

• Magnetization density: 0.096695

• e_above_hull: 0.0 eV/atom

• Curie temperature: 591.84 K

• No imaginary modes detected

• Easy axis z: -9206.405781 eV (θ=0.0°, φ=0.0°)

• Hard axis x: -9206.402854 eV (θ=90.0°, φ=0.0°)

• ΔMAE (hard-easy): 2.927303 meV

Great first result. Checks all our boxes. Slightly low on the magnetic saturation.

• Number of atoms in the cell: 4

• Magnetization density: 0.127557

• e_above_hull: 0.134595 eV/atom

• Curie temperature: 745.58 K

• No imaginary modes detected

• Easy axis z: -7302.045922 eV (θ=0.0°, φ=0.0°)

• Hard axis x: -7302.045903 eV (θ=90.0°, φ=0.0°)

• ΔMAE (hard-easy): 0.018796 meV

• Number of atoms in the cell: 16

• Easy axis z: -35245.522685 eV (θ=0.0°, φ=0.0°)

• Hard axis x: -35245.518820 eV (θ=90.0°, φ=0.0°)

• ΔMAE (hard-easy): 3.865130 meV

Not a realistic candidate but an interesting test. Iridium is extremely expensive. More expensive than Platinum. But look at the size of that MAE 👀

• Number of atoms in the cell: 4

• Easy axis x: -7645.989053 eV (θ=90.0°, φ=0.0°)

• Hard axis z: -7645.989006 eV (θ=0.0°, φ=0.0°)

• ΔMAE (hard-easy): 0.047192 meV

Unable to test! Crashed the model. Not sure what's up at this point. It's a MatterGen generation - they've been giving me all sorts of trouble recently.

• Number of atoms in the cell: 5

• Easy axis z: -9726.391907 eV (θ=0.0°, φ=0.0°)

• Hard axis x: -9726.391869 eV (θ=90.0°, φ=0.0°)

• ΔMAE (hard-easy): 0.037788 meV

• Number of atoms in the cell: 4

• Magnetization density: 0.076451

• e_above_hull: 0.184240 eV/atom

• Curie temperature: 546.18 K

• No imaginary modes detected

• Easy axis x: -7960.833456 eV (θ=90.0°, φ=0.0°)

• Hard axis z: -7960.830609 eV (θ=0.0°, φ=0.0°)

• ΔMAE (hard-easy): 2.847346 meV

This was good to see! MnBi is a known good candidate. The first non-rare material with a high MAE I've seen. Struggles on Ms and Curie but it was still a good confirmation on the model's ability to predict.

Let's see if we can take what we found in MnBi and crank up the Ms a bit:

• Number of atoms in the cell: 8

• Magnetization density: 0.159473

• e_above_hull: 0.254313 eV/atom

• Curie temperature: 637.98 K

• No imaginary modes detected

• Easy axis y: -17840.774452 eV (θ=90.0°, φ=90.0°)

• Hard axis z: -17840.771466 eV (θ=0.0°, φ=0.0°)

• ΔMAE (hard-easy): 2.985789 meV

Exciting! Good MAE and good Ms from this one. Only issue is it's formation energy is high. ehull around 0.250.

• Number of atoms in the cell: 4

• Magnetization density: 0.124042

• e_above_hull: 0.01735 eV/atom

• Curie temperature: 855.55 K

• No imaginary modes detected

• Easy axis z: -8851.951103 eV (θ=0.0°, φ=0.0°)

• Hard axis x: -8851.947816 eV (θ=90.0°, φ=0.0°)

• ΔMAE (hard-easy): 3.287396 meV

Pretty good performance. We need to figure out the role of Ni and if that is needed. All this MAE is coming from the Pt which is expensive, so we should figure out alternatives like MnBi.

Will be testing more throughout the day.